Lifetime Prediction Modeling of Automotive Proton Exchange Membrane Fuel Cells

Knowledge about the health conditions and expected lifetime of an operating fuel cell stack is essential to system control and maintenance of a fuel cell vehicle. To quickly and accurately estimate a stack’s lifetime, a data-driven prediction model for proton exchange membrane fuel cells (PEMFCs) is proposed in this study. In this model, the voltage output of the fuel cell stack is taken as the lifetime evaluation index. Two methods are used to establish the lifetime decay evaluation criteria of the PEMFC stack, i.e., (1) Least Squares Fitting (LSF) method that establishes the standard for stack voltage degradation behavior, and (2) Back Propagation (BP) neural network that learns the stack’s voltage decay characteristics and establishes the standard for the stack’s voltage degradation behavior. The Autoregressive Moving Average (ARMA) time series model is then employed to learn part of the known decay behavior of stack voltage so as to predict future stack decay. To verify the accuracy of ARMA predictions, the predicted results are compared against two established voltage decay standards as well as the real voltage output. It is found that the ARMA prediction accuracy is satisfactory. After obtaining the real cell voltage decay data within the prediction range, the lifetime prediction of the stack for the next period can be obtained using ARMA.